- From: Hamish Harvey <david.harvey@bristol.ac.uk>
- Date: Thu, 26 Aug 2004 16:23:37 +0100
- To: www-rdf-interest@w3.org
--On 26 August 2004 09:08 -0400 Dan Brickley <danbri@w3.org> wrote: >> I was simply suggesting that the approach Dan presented seems >> analogous to making assertions about the graph in which statements >> occur, from which one can infer things about those statements. > > It's v similar, except I really do make statements about the document > rather than the things written in the document (whether conceived of as > a graph, or as a bunch of statements). So the document has properties > like checksum/hash, size in bytes, etc etc., as well as authorship. > Works for me. So the statements you are making are actually about the document. Checksum, size in bytes, are definitely statements about the document, not statements about the graph serialised in the document. Authorship could be less clear, though in the case of hand authored documents authoring the document and authoring the triples become one. With a document serialised from a triple store, where triples can have different authors, this falls down; the document doesn't really have a well defined author, although it does have a well defined, say, generator. So as Steve suggested, you'd need to jump through another hoop to use this approach to make assertions about graphs without losing the ability to differentiate between graphs and documents. At which point you have invented named graphs, I think, but with the damaging restriction that a graph can only be the totality of what is contained in a single document. Still, might it be possible to do this in a way which is forward compatible with a full named graph treatment? Full named graphs just release the restriction that the only way you can actually define the graph is to put it in a document and declare that _:graph ng:obtainedByDeserialising <http://...> . Cheers, Hamish
Received on Thursday, 26 August 2004 15:25:53 UTC